[HELM] Refactor Chart

[lakinduakash]

Reference : #1871

[EshamAaqib]

@achraf-mer Just wondering if we could remove the Stack mode, is this in use ? Ideally on K8s vLLM should run separately instead of the same pod as h2oGPT I think. WDYT ?

[achraf-mer]

@achraf-mer Just wondering if we could remove the Stack mode, is this in use ? Ideally on K8s vLLM should run separately instead of the same pod as h2oGPT I think. WDYT ?

yes, we can do separate and keep the help straightforward, let's do, I think we might have used the same pod for latency considerations, but since vLLM can be resource intensive, it is best IMO to have on a separate pod. (more isolation and we can scale separately)

[EshamAaqib]

@lakinduakash Lets remove Stack mode from h2oGPT and the checks as well, similar to what was done with Agents

[lakinduakash]

@lakinduakash Lets remove Stack mode from h2oGPT and the checks as well, similar to what was done with Agents

Stack is removed

[EshamAaqib]

Lets document the breaking changes that was done to the chart, ex: changing the path of the model lock in values. We will need to communicate this to the other teams